Au@CeO2 nanozyme based smart colourimetric sensor for cholesterol: A neural network powered point of care solution model

Abstract

The current investigation aims to provide a ready-to-use biosensor with the aid of an ANN-powered point-of-care solution model for the detection of cholesterol. The synthesized AuNP@CeO₂ nanoparticles exhibited strong catalytic performance with the respective K_m (Michaelis-Menten constant) and V_max (maximum reaction velocity) values of 5.3255 mM and 0.00406 mM s⁻¹ of cholesterol. Both the visual observation and UV-Vis spectroscopy verified the colourimetric shift from blue-green to light blue. The image J results established a good linearity in the cholesterol content ranging from 0 to 50 mM, with a regression equation of y = 128.40 + 0.653x and R² value of 0.9842. Similarly, the optimal validation mean squared error (MSE) of 0.0577 has been achieved using the hard shrink activation with 13 neurons. The lowest training MSE value of 0.0637 is observed with the SELU activation function at 7 neurons. The median training and validated MSEs across all models are in the order of 0.0896 and 0.0769. Activation functions such as hard shrink, Selu, Hard Tanh, Relu6 and Tanh exhibited a uniform and consistent performance. The overall results support the effectiveness of nanozymes and ANNs for the biomedical regression tasks, especially in small sample scenarios where the capture of non-linear interactions is critical.